A review by archytas
The Feeling of Life Itself: Why Consciousness Is Widespread But Can't Be Computed by Christof Koch
4.0
Koch is an engaging, provocative writer and thinker. This book was wildly different in tone to Confessions of a Romantic Reductionist, but both have provided a pretty rich intellectual feast. I got so frustrated trying to explain this book to my partner, that I got him to read a bit, and then he read most of it, and then we talked for days about all the things we disagreed with, and where Koch is not at all convincing and so on. Frankly, we're still going a bit.
Whereas in previous work, Koch defies the dispassionate scientist stereotype, letting his passion for understanding drive the narrative, the Feeling of Life Itself is an attempt to thoroughly explain Giulio Tononi's Integrated Information Theory of consciousness. It is much more of a traditionally structured science book than I expected, and there is little of the 'big issues' that Koch does not tackle, even if much of the time I found the argument unconvincing.
IIT is a theory which argues, effectively, that consciousness is produced by the interaction of binary neurons governed by both internal and external constraints. This interaction - mappable as a series of logic gates - creates a "Whole" which is 'irreducible' to individual components, as the system is causal upon itself. This provides a measurable approach to consciousness - you can map and count the interaction in a brain while thinking (Koch and Tononi use a measure they dub Phi). Because the system has internal causality, Koch argues that it has intrinsic properties (while acknowledging this is not *exactly* a physicists definition of intrinsic, which would require those properties to exist without external stimulus).
Koch starts, as he must, with a definition of consciousness. One which is, by necessity, somewhat vague. Consciousness is experience. Koch does not really elaborate here, beyond a lot of ranting about Dennett being irritating, which is a shame, because clarifying what Koch *doesn't* include in his definition - the sense of internal narrative, and perhaps more crucially, a person's sense of self - is important for determining what causes it, and how confident we are as to whether any given entity 'is conscious'. Koch's fury at Dennis is largely because Dennett rationalises consciousness out of existence, regarding it as an 'illusion' which Koch points out denies the actual experiences of real people. While I am always in favour of having a go at Dennett - who is unapologetic is assuming philosophy will lead understanding and science will follow, which inevitably leads to the conclusion that we know nothing at all - it is also important to note that we all have very specific experiences of being conscious, and there are significant challenges in establishing how much of that specificity is shared. Consciousness in this book is treated as a binary: you have it or you don't have it. There is no discussion of how the differing structure of brains between consciousness might be a spectrum, and hence a wildly differing phenomenon. While there is some discussion about how particular areas of the brain seem involved in both phi and conscious thought - for example, the posterior cingulate cortex - there is no examination of what this might mean for species with different brain structures for experience. At one point, an equivalence between consciousness and affective empathy is thrown out there as if all species which have the first, must have the second (has the author met cats?). There was little exploration of time, despite some intriguing point at the start that hallucinogens are pretty strong evidence that a sense of time is not intrinsic to experience.
Koch presents a compelling argument that Phi is measuring *something*, but not a compelling argument that that something is experience. There is a whole chapter about the relationship between intelligence and consciousness, in which Koch shows indications that Phi correlates with species-level intelligence measures, and some interesting correlations between Phi and patients in various states of conscious impairment - which could also translate to intelligence as well as 'experience'. This is a big issue for the thesis. Phi is demonstrated as something widespread among animals, but - crucially - without being able to link this to a sense of experience, it is feels off.
Throughout the book, Koch treats consciousness as if it is a system designed to interpret reality as exactly as possible. He fails to discuss the various pieces of evidence that suggest that what we are conscious of is not only a limited section of what we are aware of, but that in some cases, it may contradict it (e.g tests that indicate that when we 'decide' to move an arm, we've already started moving it, or brain injuries in which people rationalise their flawed perceptions with events that never happened). Koch argues that we might have "multiple consciousnesses" to explain how we perform actions without being conscious of them, or how split brains work, without really explaining why the simpler solution - that conscious experiences is just one part of brain function not the (lower case) whole - is not preferred. It is hard not to think that this is because Koch wants consciousness to be a fundamental law of the universe, something fundamental, and well, sacred.
In all the book, Koch displays this desire to equate Phi to a property like gravity. One of the sillier parts of the book is when he argues that expecting computer intelligence to simulate consciousness is like expecting a Black Hole simulation to suck the science team running it into the computer. None of these analogies (note my earlier comments about the switcheroo around intrinsic) worked for me, nor did I see the point of them. Koch himself in the past has pointed out how from demonstrable/observable M-theory is. Sometimes this seems to come at the expense of examining the evolutionary purpose of conscious thought. I remain most fond of the attention theory here - that conscious thought enables us to manage considerable attentional brain resource on things of importance. Koch dismisses this in passing - actually one thing that kinda delighted me is that little of what I thought was missing was not acknowledged by Koch, it's just that his responses were unconvincing or simply "yeah, I know, going to say it anyway" in tone.
So it might seem that all I had was complaints. In the end, I was not convinced that the evidence for Phi equalling consciousness was that strong - but neither was I convinced it couldn't. I had a great time reading the book, highlighted the living hell out of it, argued about it, and have left. So. Much out of this review because it is already ridonkulous. I would love to have a rich vegan meal, including wine, with Koch and argue it all through, and if he publishes another book, I'll have it on preorder so fast your head will spin.
Whereas in previous work, Koch defies the dispassionate scientist stereotype, letting his passion for understanding drive the narrative, the Feeling of Life Itself is an attempt to thoroughly explain Giulio Tononi's Integrated Information Theory of consciousness. It is much more of a traditionally structured science book than I expected, and there is little of the 'big issues' that Koch does not tackle, even if much of the time I found the argument unconvincing.
IIT is a theory which argues, effectively, that consciousness is produced by the interaction of binary neurons governed by both internal and external constraints. This interaction - mappable as a series of logic gates - creates a "Whole" which is 'irreducible' to individual components, as the system is causal upon itself. This provides a measurable approach to consciousness - you can map and count the interaction in a brain while thinking (Koch and Tononi use a measure they dub Phi). Because the system has internal causality, Koch argues that it has intrinsic properties (while acknowledging this is not *exactly* a physicists definition of intrinsic, which would require those properties to exist without external stimulus).
Koch starts, as he must, with a definition of consciousness. One which is, by necessity, somewhat vague. Consciousness is experience. Koch does not really elaborate here, beyond a lot of ranting about Dennett being irritating, which is a shame, because clarifying what Koch *doesn't* include in his definition - the sense of internal narrative, and perhaps more crucially, a person's sense of self - is important for determining what causes it, and how confident we are as to whether any given entity 'is conscious'. Koch's fury at Dennis is largely because Dennett rationalises consciousness out of existence, regarding it as an 'illusion' which Koch points out denies the actual experiences of real people. While I am always in favour of having a go at Dennett - who is unapologetic is assuming philosophy will lead understanding and science will follow, which inevitably leads to the conclusion that we know nothing at all - it is also important to note that we all have very specific experiences of being conscious, and there are significant challenges in establishing how much of that specificity is shared. Consciousness in this book is treated as a binary: you have it or you don't have it. There is no discussion of how the differing structure of brains between consciousness might be a spectrum, and hence a wildly differing phenomenon. While there is some discussion about how particular areas of the brain seem involved in both phi and conscious thought - for example, the posterior cingulate cortex - there is no examination of what this might mean for species with different brain structures for experience. At one point, an equivalence between consciousness and affective empathy is thrown out there as if all species which have the first, must have the second (has the author met cats?). There was little exploration of time, despite some intriguing point at the start that hallucinogens are pretty strong evidence that a sense of time is not intrinsic to experience.
Koch presents a compelling argument that Phi is measuring *something*, but not a compelling argument that that something is experience. There is a whole chapter about the relationship between intelligence and consciousness, in which Koch shows indications that Phi correlates with species-level intelligence measures, and some interesting correlations between Phi and patients in various states of conscious impairment - which could also translate to intelligence as well as 'experience'. This is a big issue for the thesis. Phi is demonstrated as something widespread among animals, but - crucially - without being able to link this to a sense of experience, it is feels off.
Throughout the book, Koch treats consciousness as if it is a system designed to interpret reality as exactly as possible. He fails to discuss the various pieces of evidence that suggest that what we are conscious of is not only a limited section of what we are aware of, but that in some cases, it may contradict it (e.g tests that indicate that when we 'decide' to move an arm, we've already started moving it, or brain injuries in which people rationalise their flawed perceptions with events that never happened). Koch argues that we might have "multiple consciousnesses" to explain how we perform actions without being conscious of them, or how split brains work, without really explaining why the simpler solution - that conscious experiences is just one part of brain function not the (lower case) whole - is not preferred. It is hard not to think that this is because Koch wants consciousness to be a fundamental law of the universe, something fundamental, and well, sacred.
In all the book, Koch displays this desire to equate Phi to a property like gravity. One of the sillier parts of the book is when he argues that expecting computer intelligence to simulate consciousness is like expecting a Black Hole simulation to suck the science team running it into the computer. None of these analogies (note my earlier comments about the switcheroo around intrinsic) worked for me, nor did I see the point of them. Koch himself in the past has pointed out how from demonstrable/observable M-theory is. Sometimes this seems to come at the expense of examining the evolutionary purpose of conscious thought. I remain most fond of the attention theory here - that conscious thought enables us to manage considerable attentional brain resource on things of importance. Koch dismisses this in passing - actually one thing that kinda delighted me is that little of what I thought was missing was not acknowledged by Koch, it's just that his responses were unconvincing or simply "yeah, I know, going to say it anyway" in tone.
So it might seem that all I had was complaints. In the end, I was not convinced that the evidence for Phi equalling consciousness was that strong - but neither was I convinced it couldn't. I had a great time reading the book, highlighted the living hell out of it, argued about it, and have left. So. Much out of this review because it is already ridonkulous. I would love to have a rich vegan meal, including wine, with Koch and argue it all through, and if he publishes another book, I'll have it on preorder so fast your head will spin.